the advanced infra-red water vapour estimator (airwave...

Atmos. Meas. Tech., 12, 371–388, 2019https://doi.org/10.5194/amt-12-371-2019© Author(s) 2019. This work is distributed underthe Creative Commons Attribution 4.0 License.

The Advanced Infra-Red WAter Vapour Estimator (AIRWAVE)version 2: algorithm evolution, dataset description andperformance improvementsElisa Castelli1, Enzo Papandrea2,1, Alessio Di Roma3, Bianca Maria Dinelli1, Stefano Casadio2, and Bojan Bojkov4

1Instituto di Scienze dell’Atmosfera e del Clima, ISAC-CNR, Via Gobetti 101, 40129 Bologna, Italy2Serco s.p.a., Via Sciadonna 24–26, 00044 Frascati, Italy3Dipartimento di Fisica e Astronomia, DIFA, Universita’ di Bologna, Viale Berti Pichat 6/2, 40127, Bologna, Italy4EUMETSAT, Eumetsat Allee 1, 64295 Darmstadt, Germany

Correspondence: Elisa Castelli ([email protected])

Received: 6 June 2018 – Discussion started: 20 June 2018Revised: 16 November 2018 – Accepted: 25 November 2018 – Published: 18 January 2019

Abstract. Total column water vapour (TCWV) is a key atmo-spheric variable which is generally evaluated on global scalesthrough the use of satellite data. Recently a new algorithm,called AIRWAVE (Advanced Infra-Red WAter Vapour Esti-mator), has been developed for the retrieval of the TCWVfrom the Along-Track Scanning Radiometer (ATSR) instru-ment series. The AIRWAVE algorithm retrieves TCWV byexploiting the dual view of the ATSR instruments using theinfrared channels at 10.8 and 12 µm and nadir and forwardobservation geometries. The algorithm was used to producea TCWV database over sea from the whole ATSR mission.When compared to independent TCWV products, the AIR-WAVE version 1 (AIRWAVEv1) database shows very goodagreement with an overall bias of 3 % all over the ATSRmissions. A large contribution to this bias comes from thepolar and the coastal regions, where AIRWAVE underesti-mates the TCWV amount. In this paper we describe an up-dated version of the algorithm, specifically developed to re-duce the bias in these regions. The AIRWAVE version 2(AIRWAVEv2) accounts for the atmospheric variability atdifferent latitudes and the associated seasonality. In addi-tion, the dependency of the retrieval parameters on satel-lite across-track viewing angles is now explicitly handled.With the new algorithm we produced a second version ofthe AIRWAVE dataset. As for AIRWAVEv1, the quality ofthe AIRWAVEv2 dataset is assessed through the comparisonwith the Special Sensor Microwave/Imager (SSM/I) and withthe Analyzed RadioSounding Archive (ARSA) TCWV data.Results show significant improvements in both biases (from

0.72 to 0.02 kg m−2) and standard deviations (from 5.75 to4.69 kg m−2), especially in polar and coastal regions. A qual-itative and quantitative estimate of the main error sources af-fecting the AIRWAVEv2 TCWV dataset is also given. Thenew dataset has also been used to estimate the water vapourclimatology from the 1991–2012 time series.

1 Introduction

A key issue in assessing the climate change is the preciseknowledge of the distribution and variability of the totalcolumn of water vapour (TCWV), i.e. the vertically inte-grated atmospheric water vapour content. Actually, TCWVis closely linked to clouds, precipitation and thus to the hy-drological cycle (Allan et al., 2014). For this reason it isone of the GCOS (Global Climate Observing System) es-sential climate variables (ECVs). Since water vapour playssuch a crucial role in meteorological as well as in climato-logical aspects, it is important to gather spatial and tempo-ral thorough information about its distribution. On a globalscale, this can be achieved through the use of satellite mis-sions. In the last decades measurements from several sen-sors were used for this purpose. Among them, sensors op-erating in the microwave region, such as the Special SensorMicrowave Imager (SSM/I on board Defense MeteorologicalSatellite Program, DMSP, satellites), are used to infer the ac-curate amount of TCWV over ocean surfaces (Wentz, 1997),

Published by Copernicus Publications on behalf of the European Geosciences Union.

372 E. Castelli et. al: The Advanced Infra-Red WAter Vapour Estimator (AIRWAVE) version 2

while sensors operating in the visible and near-infrared spec-tral range provide precise TCWV retrieval on land surfaces(e.g. the ENVISAT Medium Resolution Imaging Spectrom-eter (MERIS), Lindstrot et al., 2012, or Moderate ResolutionImaging Spectroradiometer, MODIS, on the Terra and Aquasatellites, Diedrich et al., 2015).

TCWV retrievals from infrared spectral regions were per-formed from Advanced Very High Resolution Radiometer(AVHRR; Emery, 1992) measurements, using the split win-dow technique, (Sobrino et al., 1991; Li et al., 2003) andfrom MODIS (Seemann et al., 2003). TCWV retrievals frominfrared channels over land suffer from the limited knowl-edge of the temperature and the emissivity of land sur-faces (Lindstrot et al., 2014). The Along-Track Scanning Ra-diometer (ATSR, Delderfield et al., 1986) instrument serieshad as its main objective the accurate retrieval of sea sur-face temperature for climate studies. However, Casadio et al.(2016) demonstrated that it is possible to retrieve accurateand precise TCWV from its daytime and night-time mea-surements, using the ATSR brightness temperature (BT) col-lected from nadir and forward views in the channels at 10.8and 12 µm in clear-sky daytime and night-time sea scenes.The algorithm (named AIRWAVE, Advanced Infra-Red WA-ter Vapour Estimator) exploits a sea emissivity dataset andcalculations made with a dedicated radiative transfer model(RTM). A detailed description of the AIRWAVE algorithmis given in Casadio et al. (2016). The first version of theAIRWAVE TCWV dataset (hereafter AIRWAVEv1), span-ning from 1991 to 2012, is freely available from the GEWEXG-VAP website (G-VAP, 2018) in the form of monthly fieldsat 2◦×2◦ regular grid resolution from 2003 to 2008 (Schroderet al., 2018). Due to the legacy of the ATSR series, andthe fact that the radiances are a fundamental climate datasetrecord, the AIRWAVE dataset is an important resource forwater vapour studies. It is worth underlying here that AIR-WAVEv1 was developed to demonstrate the possibility of re-trieving TCWV values from the ATSR measurements. Themain goal pursued in its development was to have simplesoftware that could produce good results when compared toindependent datasets. For this reason, in the AIRWAVEv1algorithm several approximations were made. AIRWAVEv1use fixed retrieval parameters along the globe and TCWV arecorrected for viewing angle variability in nadir and slant byusing an empirical correction factor.

Papandrea et al. (2018), aiming to validate the AIR-WAVEv1 dataset, compared the data with the TCWV fromSSM/I and Analyzed RadioSounding Archive, ARSA (2018)for the whole mission. This exercise demonstrated that theAIRWAVEv1 dataset was in general of high quality (averagecorrelative bias of 0.72 kg m−2 vs. SSM/I and 0.80 kg m−2

vs. ARSA, below the 1 kg m−2 indicated in the GlobVapourproject, Lindstrot et al., 2010) apart for the polar regionsand some coastal regions, where an underestimation of theTCWV was found. In this paper we describe the new ver-sion of the AIRWAVE algorithm (hereafter AIRWAVEv2)

developed to overcome these weaknesses by accounting forlatitudinal and angular variations of the retrieval parame-ters. The new algorithm has been applied to all the avail-able ATSR level 1B top-of-atmosphere radiance products ac-quired over water surfaces in clear-sky and in day and nightconditions (same as for AIRWAVEv1) to produce the AIR-WAVEv2 dataset. Here, we show the new TCWV climatolo-gies derived from 20 years of ATSR data together with theresults of an extensive validation exercise repeating the samecomparisons reported in Papandrea et al. (2018). The newdataset shows improvements in terms of both bias and spreadof the differences with respect to another dataset and whatwas achieved with AIRWAVEv1.

This article is structured as follows: in Sect. 2 we describethe new algorithm developed to produce AIRWAVEv2, theimprovements in retrieval scenarios and the strategy used forthe selection of latitude- and seasonal-dependent retrieval pa-rameters. In Sect. 3 we describe the AIRWAVEv2 dataset,the TCWV climatology and its validation against SSM/I andARSA data and compare the performances of AIRWAVEv2and AIRWAVEv1. Finally, conclusions are given in Sect. 4.

2 AIRWAVE version 2

Papandrea et al. (2018) demonstrated the high quality ofAIRWAVEv1 by comparing the retrieved TCWV with cor-responding SSM/I and ARSA TCWV. However, in the samepaper, the authors highlighted that, at latitudes higher than50◦, the agreement was not as good as for the rest of theglobe. They speculated that this was due to the fact thatAIRWAVEv1 makes use of retrieval parameters calculatedthough RTM simulations of tropical and midlatitude atmo-spheric scenarios, which are then averaged and used for thewhole globe. This choice was driven by the considerationthat, AIRWAVEv1 being applicable to water and cloud-freescenes only, the number of cloud-free measurements over thesea at high latitudes is significantly smaller than at midlati-tudes and tropical regions. Thus, a trade-off between gen-erality (i.e. good precision at all latitudes), actual latitudi-nal coverage of cloud-free measurements and software com-plexity was the main driver for this choice. Moreover, AIR-WAVEv1 makes use of the retrieval parameters computed forthe along-track viewing geometries only and uses an a poste-riori correction for the scenes pointing outside the orbit track.

The need to have a TCWV dataset of homogeneous qual-ity at all latitudes and viewing geometries has driven thedevelopment of an improved version of the AIRWAVE al-gorithm, AIRWAVEv2. The improvements were achievedthrough three main steps. Firstly, we modified the way inwhich some of the approximations of the solving equationswere handled, leading to an improved retrieval precision.Secondly, we compute the retrieval parameters for differentlatitude bands and for 4 months that, in the retrieval, areused as look-up-tables. Finally, we calculated the retrieval

Atmos. Meas. Tech., 12, 371–388, 2019 www.atmos-meas-tech.net/12/371/2019/

E. Castelli et. al: The Advanced Infra-Red WAter Vapour Estimator (AIRWAVE) version 2 373

parameters for different viewing angles to directly accountfor across-track variations. We discuss these modifications inthe following subsections. All the computations described inthe paper have been made using the HITRAN2008 database(Rothman et al., 2009) for the spectroscopic data and theIG2 database (Remedios et al., 2007) version 4.1 for the at-mospheric scenarios. The IG2 database was developed to beused as a model atmosphere in the analysis of the ENVISATMichelson Interferometer for Passive Atmospheric Sound-ing (MIPAS) measurements (that cover the same spectral re-gion of the IR channels of ATSR). The IG2 database con-tains atmospheric vertical profiles of pressure, temperatureand abundances of the molecules active in the MIPAS spec-tral region, which are different for each year of the missionand divided into six latitudinal bands (polar, midlatitude andequatorial for the northern and southern hemispheres) andfour seasons.

2.1 Improvements in the solving equations

The starting point for the calculations of AIRWAVEv2 re-trieval parameters is the master equation of AIRWAVE ver-sion 1. Since the expressions are the same for both nadirand forward geometry, here we report the equations for thegeneral case, omitting geometries denoted by subscript NAD(for NADIR) and FWD (for FORWARD). Here we repro-duce some of the equations reported in Casadio et al. (2016)to help the reader in the comprehension of this article. Themaster equation of the AIRWAVE algorithm is Eq. (12) ofthe above-mentioned work (now Eq. 1):

lnJλ11

Jλ22

= lnFλ11

Fλ22

+ lnελ11

ελ22

+ lnγλ11

γλ22

+ λ2τ2− λ1τ1. (1)

The subscript numbers 1 and 2 represent the terms calcu-lated in the 10.8 and 12 µm channels respectively. λ1 is thevalue of the frequency in the 10.8 µm channel and λ2 is thatin the 12 µm channel. J1 is the radiance that reaches the TOAfor the 10.8 µm channel, J2 is the radiance that reaches theTOA for the 12 µm channel, F includes the atmospheric (Ja)and surface radiance (Js) contribution and is F = 1+ Ja

e−τ Js, ε

is the sea emissivity, γ is a constant arising from the Plancklaw, τ are the optical depths at the two wavelengths. Sinceonly H2O and CO2 significantly affect the optical depth intoATSR thermal infrared (TIR) channels we can write the fol-lowing:

lnJλ11

Jλ22

= lnFλ11

Fλ22

+ lnελ11

ελ22

+ lnγλ11

γλ22

+ λ2τH2O2

− λ1τH2O1 + λ2τ

CO22 − λ1τ

CO21 . (2)

By renaming,

G= lnFλ11

Fλ22

,E = lnελ11

ελ22

,χ = lnγλ11

γλ22

, (3)

we get

lnJλ11

Jλ22

=G+E+χ + λ2τH2O2 − λ1τ

H2O1

+ λ2τCO22 − λ1τ

CO21 . (4)

In AIRWAVEv1 we assume that the optical depth (τ ) isthe product of the vertical column of H2O using the relativeeffective absorption cross section (λσ ), normalised to the airmass factor (AMF) for the given line-of-sight angle:

λ2τH2O2 − λ1τ

H2O1 =

λ2σ2− λ1σ1

AMFTCWV. (5)

This equation shows that linear behaviour exists betweenthe water vapour optical depth and the TCWV. The lineardependence is exploited to solve the AIRWAVE equation andto retrieve TCWV.

In the development of AIRWAVEv2 we investigated thepossibility of finding a more accurate solution to the AIR-WAVE equation while still preserving the linear dependencebetween water vapour optical depth and TCWV. We recallhere that in AIRWAVEv1 the water absorption cross sec-tions were obtained using MODTRAN (Berk et al., 2008),while all the other values were obtained with the dedicatedRTM, developed for ATSR measurements simulations anddescribed in Casadio et al. (2016). For AIRWAVEv2, a dif-ferent approach was adopted for the calculation of effectiveabsorption cross section. We have simulated ATSR syntheticradiances for the different atmospheric scenarios of the IG2database and thus with different water vapour contents. A de-tailed description of these simulations is given in Sect. 2.2.

Using these simulations and ATSR-SSM/I-collocated

TCWV, we verified that ln Jλ11

Jλ22

correctly reproduces the real

measurement behaviour as a function of TCWV and that thisrelation is a first linear approximation.

In Fig. 2 the coloured dots represent the values of thelogarithm of the radiance ratio in Eq. (1) as a function ofthe TCWV for the different atmospheric scenarios. We re-port only the values obtained for the subsatellite scans usingthe IG2 water profiles for the summer season multiplied for0.5 and 1.5. The different colours represent different latitudebands (going from red for tropical to blue for polar). Thegrey dots represent the radiance ratio calculated from along-track AATSR measurements on the 5 and 6 August 2008 ag-gregated at SSM/I resolution (0.25◦× 0.25◦). The value ofTCWV associated with each AATSR subsatellite measure-ment was obtained from coincident SSM/I measurements. Inorder to minimise the impact of random error, only measure-ments with SSM/I pixel coverage (calculated as the ratio be-tween the actual and the maximum number of ATSR mea-surements that can be present into a SSM/I pixel) greaterthan 10 % were used for this exercise. Figure 2 shows that(a) the simulated radiances correctly reproduce the real mea-

www.atmos-meas-tech.net/12/371/2019/ Atmos. Meas. Tech., 12, 371–388, 2019


surement behaviour, and (b) the relation between the radi-ances and the TCWV can be considered linear. Actually, inthis case we find a correlation of 0.904 for real data and 0.92for the simulated ones (p value of 7.3× 10−05).

Therefore we can now rewrite Eq. (4), isolating the termsthat account for the water content (τH2O

1 and τH2O2 ):

lnJλ11

Jλ22

−G−χ −E− λ2τCO22 + λ1τ

CO21

= λ2τH2O2 − λ1τ

H2O1 =1τ. (6)

The G term of Eqs. (3) and (4) is not as constant as sup-posed, and as is partially verified in Casadio et al. (2016)and may depend on the differing water vapour content. Forthis reason, for each atmospheric scenario the average of alltheG values obtained with different water vapour contents isused (to vary the water vapour content we multiplied the wa-ter vapour profile for 0.5, 0.75, 1., 1.25 and 1.5). Equation (6)can thus be written as follows:

lnJλ11

Jλ22

−GAVG−χ −E− λ2τCO22 + λ1τ

CO21 =1τ. (7)

Therefore for each scenario and geometry we can write

lnJλ11

Jλ22

−GAVG−χ −E− λ2τCO22 + λ1τ

CO21

=1τ =1σ ·TCWV+1ρ. (8)

In this equation, 1σ and 1ρ represent the slope and in-tercept of the straight line representing the behaviour of theterm containing the radiances as a function of the TCWV. Inthe testing version of the AIRWAVEv2 code, we estimatedthese parameters using the values of the radiances and theTCWV obtained perturbing the IG2 water vapour amount byfactors of 0.5 and 1.5. By grouping the terms in Eq. (8) as inCasadio et al. (2016) we get

TCWV=8−G

1σ, (9)

where 8 is the “water vapour pseudo-column” that in AIR-WAVEv2 is defined as follows:

8=

ln Jλ11

Jλ22

−χ −E− λ2τCO22 + λ1τ

CO21 −1ρ

1σ. (10)

This formula is slightly different from the one used inCasadio et al. (2016) due to the presence of the 1ρ term:

8=

ln Jλ11

Jλ22

−χ −E− λ2τCO22 + λ1τ

CO21

1σ. (11)

If in Eq. (9) now we make explicit the dependence on theviewing angles, we get the following:

TCWVNAD =8NAD−GNAD

1σNAD(12)

and TCWVFWD =8FWD−GFWD

1σFWD, (13)

with

8NAD =

lnJλ11NAD

Jλ22NAD

−χ −ENAD− λ2τCO22NAD

+λ1τCO21NAD−1ρNAD

1σNAD

(14)

8FWD =

lnJλ11FWD

Jλ22FWD

−χ −EFWD− λ2τCO22FWD

+λ1τCO21FWD−1ρFWD

1σFWD

. (15)

The TCWV can be estimated through knowledge of G.Actually, for single-view geometry, the presence of the Gterm can affect the accurate determination of TCWV. Thus,the G variability suggests that it will be desirable to avoidthis term in TCWV derivation. In AIRWAVE this is doneby exploiting the dual-view capability of the ATSR instru-ments and by assuming a perfect collocation between NADand FWD measurements:

TCWV= α ·8NAD+β ·8FWD, (16)

where

α =1

1− 1σFWDδ·1σNAD

and β =1

1− δ·1σNAD1σFWD

(17)

with δ ≈GFWD

GNAD. (18)

Equations (10), (12) and (16) are the solving equationsused in AIRWAVEv2, while AIRWAVEv1 makes use ofEqs. (11), (12) and (16). Equations (10), (12) and (16) weresolved for the 11 couples of viewing angles correspondingto the tie points. The angles cover a range from 0 to 21◦ inthe NAD case and from 53 to 55◦ in the FWD case. Thenew equations were used to compute a new set of retrievalparameters. For consistency purposes, in AIRWAVEv2 theywere computed with the dedicated RTM as they were forAIRWAVEv1. An example of the difference between the pa-rameters used for AIRWAVEv1 and AIRWAVEv2 is given inTable 1 for the tropical scenario and subsatellite view con-figuration. As can be noticed, the larger differences are forα, β and 1σ parameters: there is a reduction in the α andβ parameters by a factor of about 50 from AIRWAVEv1 toAIRWAVEv2, while 1σ is reduced by a factor of 30.

These changes have a direct effect on the retrieval preci-sion. We can have an estimate of the improvements of theprecision of AIRWAVEv2 retrievals against AIRWAVEv1 us-ing the following consideration: in AIRWAVE we can es-timate the expected precision by multiplying the measure-ment random error by a factor of α/1σ . Therefore, inAIRWAVEv1, we multiply the random error by a factor of



Table 1. AIRWAVE v1 and AIRWAVEv2 retrieval parameters for tropical summer atmosphere and along-track configuration (subsatelliteview at 0 and 55◦).

α β δ 1σNAD 1σFWD GNAD GFWD 1ρNAD 1ρFWD10−06 10−06 10−06 10−06 10−06 10−06

ATSR-1 AIRWAVEv1 50.7 −49.7 1.65 1.49 2.41 −5.30 −8.82 — —AIRWAVEv2 1.72 −0.72 1.69 0.06 0.05 −6.59 −11.1 6.59 11. 2

ATSR-2 AIRWAVEv1 50.5 −49.5 1.63 1.74 2.78 −6.36 −10.4 — —AIRWAVEv2 1.65 −0.65 1.67 0.07 0.04 −7.88 −13.1 7.89 13.2

AATSR AIRWAVEv1 53.1 −52.1 1.62 1.90 3.02 −7.06 −11.5 — —AIRWAVEv2 1.67 −0.67 1.66 0.08 0.05 −8.74 -14.5 8.77 14.6

53/1.9(= 28), while in AIRWAVEv2 the multiplicative fac-tor is 1.7/0.08(= 21). In the case of polar atmosphere thisratio is further reduced, reaching a value of 13. Since theoverall random error is about 0.25 % for AATSR and ATSR-2and 0.6 % for ATSR-1, the precision for AATSR and ATSR-2 improves from 7 % to 5 % and for ATRS-1 from 17 % to12 % for tropical atmosphere. In AIRWAVEv2 it reaches 3 %for AATSR and ATSR-2 and 8 % for ATSR-1 in polar at-mosphere. In the worst cases, the precision has at maximuma 1 % higher value in the extreme across-track of the swathwith respect to the subsatellite points. From these considera-tions it follows that AIRWAVEv2 parameters should improvethe retrieval.

2.2 Improvements in the retrieval scenario

In AIRWAVEv1 we make use of the same set of retrievalparameters for all measurements of a single ATSR instru-ment. Independent sets of parameters are calculated for thethree missions, while within each mission no dependencieson different atmospheric–surface conditions or seasons wasconsidered. In AIRWAVEv2 the retrieval parameters are es-timated, not only according to the instrument type but alsoaccounting for possible latitudinal and seasonal variations.To this aim, we have used the aforementioned RTM to com-pute all the required quantities by exploiting the model at-mospheres of the IG2 database. Since the IG2 database wasspecifically developed for the ENVISAT MIPAS mission, itcovers only the time range from 2002 to 2012, while theATSR series operated from 1991 to 2012. However, the inter-annual variations of most of the species active in the ATSRthermal infrared spectral range are generally much smallerthan the corresponding seasonal ones. Therefore we usedthe data for 1 year only (2010), and we considered the in-terannual variations to be systematic error sources (see Ap-pendix A for an estimate of these errors).

To better reproduce the variability of the atmospheric sce-narios that were observed by the ATSR instruments, wetherefore calculated the retrieval parameters exploiting theprofiles for all six latitude bands and the four seasons in-cluded in the IG2 datasets for 1 year only.

The AIRWAVE solving equations do not explicitly makeuse of the sea surface temperature (SST). However, SST isused in the radiative transfer (RT) computations to estimatethe retrieval parameters. The sea surface emissivity valuesare instead used both in the parameter estimation and in theAIRWAVE retrieval. While over land the emissivity is char-acterised by a large spatial difference (it indeed varies as afunction of soil type, vegetation cover, etc.), over sea its vari-ation is in general relatively small. For this reason, in AIR-WAVEv1, we used constant emissivity values calculated forthe nadir and forward viewing angles with fixed SST (285 K)and wind speed (3 m s−1). In the new version of the algo-rithm, coherently with the approach used for the atmosphericscenarios, the retrieval parameters have been computed usingdedicated SST values for each season and latitude band. TheSST monthly means were produced for the corresponding sixlatitude bands and for the four seasons using ECMWF ERA-Interim daily field data with a regular latitude–longitude gridof 0.75◦× 0.75◦ (241× 480 grid points).

The emissivity of each scenario has been computed us-ing data extracted from the University of Edinburgh database(Embury et al., 2008). This dataset contains emissivities tab-ulated as a function of wave number (600–3350 cm−1 or 3–16.7 µm), viewing angle (0–85◦), temperature (270–310 K),and wind speed (0–25 m s−1 at 12.5 m). For the RTM compu-tations we used the full spectral dependency of the emissiv-ity. Since in Eq. (3) of AIRWAVE we use a single emissivityvalue for each channel, we estimated it by convolving thespectral emissivity with the ATSR filter functions. The nadirand forward viewing angles of the instruments have been de-fined at 11 tie points of the ATSR swath (pixels associatedwith specific points equally spaced across a single image orinstrument scan). For each tie point we then used the corre-sponding viewing angles to extract the correct emissivity val-ues, with a fixed wind speed (3 m s−1), as for AIRWAVEv1.Due to wind speed variability, in both algorithm versions, weprefer to fix the value of the wind speed and then treat windvariations as an error source (see Sect. A3 in Appendix).



2.3 Across-track variations of the retrieval parameters

A simplification present in AIRWAVEv1 is that the retrievalparameters are calculated only for the subsatellite viewingangles (55◦ for forward viewing and 0◦ as nadir viewingangle). However, due to the ATSR configuration, the nadirviewing angles vary from 0◦ (subsatellite) to approximately21◦ (across-track edge of the ATSR swath, ±250 km fromnadir), while the forward viewing angles range from 53 to55◦. Significant TCWV differences between the centre andedge swaths are thus expected.

In AIRWAVEv1 the across-track dependence of TCWV iscorrected a posteriori. The correction was calculated on thebasis of TCWV retrievals performed over simulated bright-ness temperatures (BTs).

Figure 1 shows the absolute difference between TCWVfrom across-track pixels and subsatellite pixels calculated forAATSR in different atmospheric scenarios (tropical and mid-latitude) from synthetic measurements. To produce these re-trievals, we simulated ATSR radiances at the 11 tie pointsfor the nadir and forward views while also considering thesurface emissivity variations with viewing angles. The atmo-spheric scenario, the TCWV and the SST were kept constantand they were exactly the same as that used for the sub-satellite track case. Then we retrieved the TCWV for eachof the 11 couples of BTs and computed the difference withrespect to the value obtained at the subsatellite track posi-tion (which coincides with the TCWV reference value). Thedotted line in Fig. 1 mimics the correction term adoptedwithin the AIRWAVEv1 algorithm. As can be seen, the dot-ted line reproduces the general behaviour of the across-trackTCWV dependence well. This is confirmed by the com-parison made between AIRWAVEv1 TCWV and SSM/I orECMWF TCWV (Casadio et al., 2016): on average we didnot find any across-track bias, thus the general validity ofthis correction is confirmed.

However, as can be seen in Fig. 1, this approximationmight not be sufficiently adequate depending on the used at-mospheric scenario. Furthermore, a slight asymmetry withrespect to the subsatellite track position is expected as theATSR instruments are tilted about 4◦ respect to the flyingdirection of the satellite. Therefore the a posteriori correc-tion of AIRWAVEv1 cannot fully reproduce all these fea-tures. In AIRWAVEv2 we replaced the a posteriori correc-tion: we calculated the retrieval parameters for each of theabove-described tie points of the nadir and forward swaths,then we obtained the parameters at the exact ground pixel po-sition interpolating these values and using the ground pixelacross-track position.

2.4 Selection of the retrieval parameters

The computation of the retrieval parameters, described in theprevious sections, produced a set of 1584 retrieval parame-ters for each ATSR mission (6 coefficients ×6 latitude bands

Figure 1. Differences between TCWV calculated from across pix-els and TCWV calculated from the subsatellite track as a function ofacross-track position calculated for different atmospheric scenariosand for AATSR (symbols). Grey dashed line corresponds to AIR-WAVEv1 parameterisation.

Figure 2. Logarithm of radiance ratio at nadir as a function ofTCWV in simulated atmospheric scenarios (purple and blue forpolar, light and dark green for midlatitude and light and dark redfor equatorial). Grey dots represent the same quantity using realAATSR radiances and coincident SSM/I TCWV for the along-trackmeasurements on 5 and 6 August 2008.

×4 seasons ×11 tie points) stored in dedicated look-up ta-bles. In order to select the most suitable set of parameters foreach ATSR measurement, a multivariate interpolation on a3-dimensional grid (trilinear) has been applied to the six re-trieval coefficients (1σ , G, 1ρ) and the emissivity for boththe FWD and the NAD geometries.



3 AIRWAVEv2 dataset: description, climatology andvalidation

The AIRWAVEv2 TCWV data are produced in the same wayas the AIRWAVEv1 dataset, by processing level 1B measure-ments acquired over water surfaces (sea and lakes) and inclear-sky conditions in both nadir and forward views (ac-cording to the level 1B cloud mask). The output files aresaved in Interactive Data Language (IDL) binary files (.savextension); however they can be easily converted to otherformats (e.g. netcdf) upon request. The parameters containedinto the files are structured in two groups: in the first one(named HIRES) the parameters are given at native resolution(1× 1 km2), while in the second one (named SSM/I) the pa-rameters are aggregated to SSM/I resolution (0.25◦× 0.25◦

grid). Both groups contain the TCWV, the latitude, the lon-gitude, the across-track index value (0–512) and a day–nightflag. The SSM/I group, in addition, contains the value of thenumber of elements aggregated within the SSM/I grid celland the standard deviation of the TCWV value associatedwith each cell.

The climatologies have been derived using all the avail-able years and sensors of the ATSR family. Using the AIR-WAVEv2 products aggregated on the 0.25◦× 0.25◦ grid, weobtained for each month day and night TCWV averages andstandard deviations. The final monthly files are also saved asIDL binary files but can be converted to other formats andare available upon request.

Figures 3 and 4 show sketches of the climatology for Jan-uary, April, July and October obtained from 20 years ofATSR daytime measurements. Similar results are obtainedfor night-time retrievals (not shown here). In Fig. 3 we reportin panels (a), (c), (e) and (g) the TCWV global distributionat 0.25◦ grid resolution and its standard deviation in panels(b), (d), (f) and (h). In Fig. 4 panels (a), (c), (e) and (g) weshow the TCWV meridional mean together with its standarddeviation, while on panels (b), (d), (f) and (h) we report theTCWV zonal mean.

The geographical distribution of the median values of theTCWV reflects the behaviour of general atmospheric circu-lation. Higher TCWV values, associated with strong convec-tion in the Intertropical Convergence Zone (ITCZ), are lo-cated around the equator, while lower values are found inthe polar regions. Also, the zonal means as a function oflatitude reflects this behaviour, while the zonal meridionalmeans show a more homogeneous behaviour. Here lowerTCWV values are found in coincidence of longitudes wherewe have extended land presence in the equatorial region. Themean and the absolute standard deviation show similar fea-tures, with higher values in the ITCZ region. As can be no-ticed, the seasonal movements of the ITCZ from the north (inNorthern Hemisphere summer) to south (in Northern Hemi-sphere winter) can be clearly detected (Castelli et al., 2018).In general, the standard deviations in the region where theTCWV maximum is located are of the order of 15 % and up

to 20 % in polar regions. The zonal means reflect the shiftof the ITCZ during the year, reaching maximum values ofTCWV in Northern Hemisphere in July.

The quality of the AIRWAVEv2 dataset is evaluatedthrough the same method adopted for AIRWAVEv1 and re-ported in Papandrea et al. (2018). The AIRWAVEv2 datasetis compared to the TCWV obtained from the SSM/I satelliteand to data available from the ARSA.

In this contest, these two datasets are complementary, asthe SSM/I TCWV are not retrieved for measurements inproximity to coasts (minimum distance of about 60 km),while the selected ARSA stations are located in coastal ar-eas.

The SSM/I dataset has been produced from the SSM/Iinstrument series on board the DMSP polar orbit satellitessince 1987. For this comparison, we used the 0.25◦ v7daily product obtained from the F13 satellite produced byremote-sensing systems. In fact, in contrast to the otherDMSP satellites, the local time of the ascending node of F13(18:00 UTC) is more stable than the one of the other satelliteswith only a variation of 1 h during the whole mission. For thecomparison with SSM/I we used the AIRWAVEv2 data ag-gregated at 0.25◦ resolution, covering the time period from1995 to 2009. The ARSA dataset spans from January 1979 tothe present and contains water vapour concentration profilesat specific pressure levels. To obtain the TCWV we verticallyintegrate these values. For the comparison with the ATSRdata, only stations surrounded, even partially, by water areused. More details about the selection of ATSR and ARSAcoincident data are reported in Papandrea et al. (2018).

The zonal means (calculated in bins of 2◦ for all thedatasets and reported in Fig. 5) show that AIRWAVEv2 dataare of high quality with respect to both radiosondes and satel-lite data at all latitudes. In Fig. 5, the AIRWAVEv1 data areoverplotted for comparison. The improvement in the perfor-mance of the new dataset is clearly visible at all latitudes,and in particular for regions at latitudes higher than 45–50◦,where the negative values obtained with AIRWAVEv1 dis-appear. In addition, a significant reduction of the spread ishighlighted.

In Fig. 6 we show the bidimensional histograms of thecomparisons between AIRWAVEv2 and SSM/I (panel a) andARSA (panel b). The SSM/I measurements are homoge-neously distributed over the globe, while ARSA radiosound-ing stations are mainly located at midlatitudes (see Fig. 7)and this is reflected in the bulk of the ARSA TCWV valuesranging between 0 and 30 kg m−2.

Globally, a good correlation is obtained from the datasets,as highlighted by the correlations (0.948 with SSM/I and0.918 with ARSA) and bias values (0.02±4.79 kg m−2 withrespect to SSM/I and 0.19± 6.12 kg m−2 with respect toARSA). We highlight that, in the validation exercise, wecompare SSM/I data in coincidence with AIRWAVEv2 ones.Since AIRWAVE is applicable only to clear-sky measure-ments, this is a method to filter out SSM/I cloudy TCWV



Figure 3. Climatology of daytime TCWV from the AIRWAVEv2 dataset for January (a), April (c), July (e) and October (g) from 1991 to2012, and standard deviations for the same months (b, d, f, h).



Figure 4. TCWV distribution as function of latitude (median and standard deviation) (a, c, e, g) and as function of longitude (b, d, f, h).

Figure 5. Zonal means of TCWV for AIRWAVEv1 (blue), AIRWAVEv2 (red) and correlative measurements (black): SSM/I (a) or ARSA (b).The data have been averaged in 2◦ latitude bins. AIRWAVE TCWV standard deviations are also reported.

and thus to avoid biases due to different sensitivities relatedto the used spectral range. When comparing AIRWAVE datato radiosondes, the small bias we found demonstrates thatAIRWAVE TCWV are also sensitive to low atmospheric lev-

els. In particular, the comparison with the same histogramsof AIRWAVEv1 (see Fig. 1 of Papandrea et al., 2018) high-lights the correction of negative values in polar and coastalregions in the new version of the dataset.



Figure 6. AIRWAVE TCWV vs. SSM/I TCWV (a) or ARSA TCWV (b). The bin size is 2.5 kg m−2. The colour scale indicates the numberof elements of the histogram. The data cover the period from 1991 to 2012.

Figure 7. Average absolute (a) and relative (b) TCWV differences (SSM/I-AIRWAVE) at 0.25◦× 0.25◦ spatial resolution. Average ARSA-AIRWAVE TCWV differences are overplotted with circles; the size of the symbols is proportional to the number of matches (see legend).The data cover the period from 1991 to 2012.

Figure 7 reports the geographical distribution of the meanTCWV differences with respect to SSM/I and ARSA in ab-

solute and percentage values. In comparison to AIRWAVEv1(see Fig. 3 of Papandrea et al. (2018)) the differences with



Table 2. AIRWAVE TCWVs compared with SSM/I and ARSA stations. Results are also given for AIRWAVEv1. Absolute differences alongwith the standard deviations are reported for the global (all latitudes), equatorial (25◦ S–25◦ N), midlatitude (25–60◦ S and 25–60◦ N) andpolar (> 60◦ N or > 60◦ S) scenarios. Average values for ATSR-1, ATSR-2 and AATSR are also provided.

SSM/I-AIRWAVE ARSA-AIRWAVE

Instrument Scenario N × 105 BIAS-v2 SD-v2 BIAS-v1 SD-v1 N × 105 BIAS-v2 SD-v2 BIAS-v1 SD-v1kg m−2 kg m−2

All Global 3110 0.02 4.69 0.72 5.75 3.01 0.19 6.12 0.80 7.73All Equator 1560 −0.17 4.79 −0.17 5.57 0.87 −0.70 6.60 −2.40 7.74All Midlatitude 1380 0.07 4.84 1.12 5.89 1.80 0.49 6.10 1.69 7.44All Polar 170 1.32 3.51 5.55 5.14 0.35 0.86 4.59 4.12 6.47ATSR−1 Global 190 −0.20 5.17 1.15 6.17 0.48 −0.71 6.24 0.23 7.62ATSR−2 Global 1390 0.24 4.77 0.80 5.87 1.00 0.70 6.04 1.13 7.65AATSR Global 1520 −0.16 4.70 0.58 5.77 1.53 0.13 6.11 0.75 7.81

SSM/I are reduced at all latitudes. The longitudinal patternsof the differences are similar to the ones of AIRWAVEv1,except for the equatorial pacific region where AIRWAVEv2shows a slightly higher positive bias. The reasons for this be-haviour are under investigation. In the majority of coastal re-gions, where AIRWAVEv1 underestimated the TCWV, AIR-WAVEv2 is now in agreement with ARSA results (no SSM/Idata close to the coast).

In Table 2 we summarise the results of this compari-son for both ARSA and SSM/I and for different scenariosand missions. The average bias is about 0.0± 4.7 kg m−2

with respect to SSM/I and 0.2± 6.1 kg m−2 with respect toARSA. If we compare these results with the ones for AIR-WAVEv1 (reported in Table 2 to ease the comparison) we canclearly see the improvement in both the biases and the stan-dard deviations (0.7± 5.7 kg m−2 with respect to SSM/I and0.8±7.7 kg m−2 with respect to ARSA). As can be seen fromFigs. 5 and 7 and from the results in Table 2 the improvementin the bias is obtained at all latitudes; it is, however, more ev-ident in polar regions (from 5.5±5.1 kg m−2 in AIRWAVEv1to 1.3± 3.5 kg m−2 in AIRWAVEv2 versus SSM/I and from4.1±6.5 kg m−2 in AIRWAVEv1 to 0.9±4.6 kg m−2 in AIR-WAVEv2 when using ARSA). Slight differences between thethree ATSR missions are consistent with the related uncer-tainties.

In Fig. 8 we show the monthly mean evolution of the dif-ferences (and their standard deviation) between the TCWVobtained from correlative measurements and AIRWAVEv2.As for AIRWAVEv1, these differences are quite stable overtime, with the exception of the beginning of the ATSR-1 mis-sion (1991–1994). As explained in Papandrea et al. (2018),this could be due to the failure of a 3.7 µm channel that im-pacted the ATSR cloud screening. In general, the differenceswith respect to the radiosonde exhibit a higher seasonalitydue to pronounced variability of atmospheric and surfaceconditions in coastal areas.

It is worth noticing that the spread of AIRWAVEv2 is al-ways smaller than the one of AIRWAVEv1. This is partially

due to the new algorithm, which reduces the random errorcomponent due to noise in the retrieved TCWV.

The above-described results indicate that the AIRWAVEv2algorithm reduces the global bias with respect to SSM/I,from about 0.7 kg m−2 of AIRWAVEv1 to 0.0 kg m−2 andimproves the standard deviation (SD) of up to 20 % with re-spect to AIRWAVEv1. In AIRWAVEv2, the SD values areessentially constant for all the scenarios and missions, high-lighting the unbiased nature of the dataset with respect toSSM/I (cloud-free). When using ARSA data the bias reducesfrom 0.8 to 0.2 kg m−2 and the standard deviations by 21 %.

4 Discussion and conclusions

The second version of the AIRWAVE TCWV dataset de-scribed in this work has been validated against ARSA andSSM/I equivalent products.

As expected also from the analysis of synthetic retrievals,the most significant AIRWAVEv2 improvement is achievedat polar latitudes. In polar regions the bias versus SSM/Iimproves of 4.2 and of 3.2 kg m−2 versus ARSA. In bothcases the standard deviations are reduced of about 1.6–1.9 kg m−2. However, improvements at midlatitudes are alsofound. The average bias with respect to SSM/I improves byabout 0.7 kg m−2 and the standard deviation is reduced byabout 1 kg m−2. In the case of validation against radioson-des, the bias in AIRWAVEv2 is reduced by about 0.6 kg m−2

with respect to AIRWAVEv1 and the standard deviation isreduced by 1.6 kg m−2.

These improvements are due to two factors. First the AIR-WAVEv2 retrieval parameters now account for the atmo-spheric variability. Secondly the implementation of the newalgorithm explicitly takes into account the geometry and lat-itude dependence of each pixel, allowing possible artefactsdue to approximations and a posteriori corrections to be over-come. No statistically significant trend can be found in thecomparison with SSM/I and ARSA in both versions of thedatabase, while a seasonal dependence of the differences is



Figure 8. SSM/I-AIRWAVE TCWV (orange) and ARSA-AIRWAVE TCWV (red) monthly mean trends. The difference between the correl-ative measurements and AIRWAVE TCWV±SD is also reported.

observed, with a larger bias in July and August, mainly dueto the differences in midlatitude north TCWV retrievals. Ingeneral, we find slightly drier results with respect to ARSAand SSM/I with both versions. This is possibly due to thefact that the temporal mismatch between the ATSR and thecorrelative measurements does not allow all SSM/I TCWVretrieval obtained under cloudy conditions to be excluded orcloud masks to be wrongly assigned. As discussed, the use ofretrieval parameters that are calculated in conditions differ-ent from the ones present in the observed scenario can causebiases in the obtained TCWV. We point out that the majorsource of errors in the retrieved TCWV comes from the tem-perature profile assumptions, while erroneous assumptions ofother gases (e.g. HNO3, CFC-11, CFC-12, CO2) have an al-most negligible impact. The obtained RMSE value of about7 % is of the same order of this error. The error quantificationgiven in this work allows the users to get a better insight ofthe AIRWAVEv2 TCWV dataset and its related quality.

Besides the improvements on ATSR TCWV retrievalsgiven by the AIRWAVEv2 dataset, the method described inthis work can be the basis for a similar approach for SLSTR(Sea and Land Surface Temperature Radiometer, on boardCopernicus Sentinel-3, Donlon et al., 2012).

Data availability. The AIRWAVEv2 data are available on requestfrom the corresponding author ([email protected]).



Appendix A: Evaluation of the main systematic andrandom errors in AIRWAVE version 2

This appendix provides an estimate of the main error sources(random and systematic) that affect the AIRWAVEv2 dataset.To facilitate the use of the errors together with the TCWVcontained in the dataset, we summarise the results of thisAppendix in Table A1. AIRWAVEv2 retrieval parameters ac-count for atmospheric and surface variability of the observedscenes. However, real observed scenes can obviously devi-ate from the simulated ones. An evaluation of the errors in-duced by these deviations is then required. For this reasonwe analyse, through the use of synthetic radiances, the majorsources of errors that can affect the AIRWAVEv2 retrieval,i.e. (in order of importance) the influence of the atmospherictemperature profile variations, the impact of sea emissivitychanges due to the wind and the impact of interfering atmo-spheric species. Furthermore, an estimate of the random errorcomponent due to the noise is also reported.

A1 Retrieval approximations

In order to evaluate the impact of systematic errors dueto the retrieval approximations, we ran some tests ofTCWV retrievals from simulated radiances. We used top-of-atmosphere (TOA) subsatellite track brightness temperatures(BTs) simulated with the RTM at 10.8 and 12 µm in nadir andforward geometries as input to the retrieval chain (no randomnoise was added). The BTs were produced using differentTCWV amounts for each given scenario (e.g. water vapourprofiles were multiplied by 0.5, 0.75, 1.25 and 1.5), whileall the other atmospheric profiles were kept constant. The re-sults of this exercise show that TCWV is correctly retrievedwith an error of ±3 %. For comparison purposes, we per-formed a similar test using the AIRWAVEv1 approach. ForAIRWAVEv2 tests, the same atmospheric conditions used tocompute the retrieval parameters have been adopted, whilefor AIRWAVEv1 larger errors are expected due to the factthat the atmospheric variability is not taken into account.Despite this, this analysis shows that AIRWAVEv1 performswell for medium-high TCWV (> 30 kg m−2), where the dif-ferences are below 10 %. For TCWV values between 10–20 kg m−2, the differences range between−10 % and−30 %.AIRWAVEv1 underestimates the low TCWV values (below10 kg m−2) with differences from −50 % to −150/−200 %.This was also reported in Casadio et al. (2016), where the au-thors compared AIRWAVE TCWV with collocated ECMWFcounterparts, showing a dry bias at high latitudes (where theTCWV values are small). The tests on synthetic radiancesindicate that the AIRWAVEv2 parameterisation solves thisissue.

Figure A1. CO2, HNO3, CFC-11 and CFC-12 spectral lines andATSR filter functions (arbitrary units).

A2 Atmospheric temperature and water vapourprofiles and SSTs

One of the main error sources that can affect the AIRWAVETCWV retrieval is the assumption of a fixed temperature pro-file. Actually, the atmospheric opacity is closely linked tothe atmospheric density and thus to the atmospheric temper-ature. To estimate the impact of temperature on the retrievedTCWV, 20 different temperature profiles were randomly per-turbed by ±3 K on a 1 km equispaced altitude grid. In orderto also account for possible changes related to SST variationwe change the SST accordingly to the value of the tempera-ture in the lowest layer. Then, simulated BTs were producedwith the RTM and were used to perform the TCWV retrievalsfor the three instruments in equatorial, midlatitude and polarJuly conditions for the Northern Hemisphere and the resultswere compared with respect to the unperturbed case. In thethird column of Table A1 we summarise the findings of thisanalysis for each of the three instruments, reporting the SDof the difference both in absolute and in percentage values.The impact of these perturbations is of the order of 6 % and ishigher in the equatorial and midlatitude regions and lower atthe poles. Indeed, in the equatorial region, due to the higherwater vapour content, the atmosphere is more opaque than atthe poles so that temperature variations have a larger impacton the retrieved TCWV. These tests were also run by vary-ing the atmospheric profile alone. Very similar but slightlyhigher errors are found in these cases. Another relevant errorsources can be due to differences in the water vapour profileshape. To evaluate this error, we varied the water vapour pro-file randomly up to 5 % at each atmospheric level. At max-imum the impact is 1 % in the equatorial case (atmosphericopacity; see the fourth column in Table A1).



A3 Wind speed

A further source of error that can affect the TCWV retrievalsis the value used for the sea emissivity, which directly enterEqs. (10) and (11). Sea emissivity depends on wavelength,sea surface temperature, viewing angles and wind speed. Asstated in Sect. 2.2, in AIRWAVEv2 we accounted for emis-sivity variations due to the viewing angles and sea surfacetemperatures. All the calculations were made at a fixed windvalue of 3 m s−1. In order to assess the possible systematic ef-fects due to wind variations on the retrieved values, we variedthe emissivity according to the wind speed at three values astabulated in the University of Edinburgh emissivity database:1, 10 and 25 m s−1.

The emissivity has a different value and spectral behaviourwith different wind speeds. The fourth column of Table A1reports the difference in the TCWV retrieved for simulatedmeasurements with the 25 m s−1 wind speed with respect tothe reference case. As expected, the impact is almost negli-gible in the equatorial band, where the higher opacity of theatmosphere reduces the sensitivity of ATSR measurementsto the surface conditions, while increasing it toward the poleswhere, due to the low atmospheric opacity, the surface effectsbecome relevant with respect to the atmospheric component.Furthermore, in the case of polar conditions, the effects of thewind on the retrieved TCWV are not linear, with enhancedvariations for wind speed of 25 m s−1, as not only the in-tensity but also the spectral shape of the surface emissivityvaries in function of the wind speed. The possible presenceof white caps, generated by high speed winds, has not beenconsidered in this study.

A4 Interfering atmospheric constituents

The AIRWAVE algorithm, in both versions, accounts for thecontribution to the radiance of the two main gases active inthe ATSR channels (H2O and CO2). However some otherspecies have spectroscopic features in the ATSR channels.In Fig. A1, the CO2, HNO3 and CFCs spectra in the 10–13 µm wavelength range are shown, along with the ATSR fil-ter functions (all in arbitrary units). In order to have a com-plete view of the possible error components, we assessed theimpact of interfering species in case their abundance differsfrom the one used in the reference scenarios to compute theretrieval parameters.

Among the considered species, HNO3 shows significantlatitudinal and seasonal variability, while CO2 and CFCs ex-hibit interannual trends (Remedios et al., 2007). For this rea-son we separately accounted for the effects due to latitudinaland interannual variability. We used the IG2 database ver-sion 4.1 and our RTM to produce synthetic BTs. For eachseason and latitude band we generated synthetic BTs usingthe proper IG2 atmospheric status, except for the profile ofthe investigated species, for which we used all the differentavailable profiles. The generated BTs were then used to re-

trieve the TCWV to assess the systematic error induced bythe expected variability of the interfering species.

Latitudinal and seasonal variations of HNO3 impactATSR-1 and ATSR-2 BTs more than AATSR, because ofthe different shapes of the TIR filter functions, and can pro-duce differences up to 0.3 K in the 11 µm band (midlatitudevs. tropical north in January). For CFC-11 seasonal differ-ences in the tropics are of the order of 0.001 K in the 12 µmband and for CFC-12 we get 0.03 K in the 11 µm band, whilelatitudinal variations reach 0.04 K from tropical to midlati-tude atmospheres for both channels. CO2 latitudinal varia-tions can produce a maximum difference of 0.003 K on nadirBT. The impact of maximum latitudinal variations of HNO3,CFCs and CO2 on the retrieved TCWV is reported in Ta-ble A1. The largest contribution is due to HNO3 latitudinalvariation and is of the order of 0.6 %, while the CFCs lati-tudinal variations produce differences of 0.15 % and CO2 of0.01 % only. Furthermore, the impact of using mid latitudeprofiles instead of tropical profiles for all species but H2Ohas a maximum impact of 0.22 %. Seasonal variations are al-most negligible for CO2 and CFCs, while they are only 0.6 %for HNO3. Thus we can safely assume that the latitudinaland seasonal variations of the interfering species represent aminor error source for the AIRWAVE TCWV for both AIR-WAVEv1 and AIRWAVEv2.

To also evaluate the impact of interannual variations for allthe ATSR series, we would need the CO2 and CFCs profilesfrom 1991 to 2012. However, as mentioned in the previoussections, the IG2 database contains data from 2002 onwards.In this work, the CFCs and CO2 profiles from 1991 to 2001were inferred by scaling the 2002 profiles using the trendgiven in the last IPCC report for CO2 and on the Mauna Loaobservatory website (Global Monitoring Division, 2018 web-site; see also Aoki et al., 2003 and Minschwaner et al., 2013)for CFC-11 and CFC-12. The impact on the ATSR BTs dueto the interannual variations of CO2 and CFCs has been eval-uated for each mission. We calculated the synthetic spectrausing the CFC-11, CFC-12 or CO2 profile for the initial andfinal year of each mission. Then we calculated, for each in-strument, the differences in retrieved TCWV at the beginningand at the end of each mission. The results for CO2 and CFCsare shown in Table A1. The influence of CO2 annual varia-tions on the retrieved TCWV is of the order of 0.004 kg m−2

(< 0.01 %). For CFCs we obtain 0.002 kg m−2. We can thenconclude that the impact of VMR latitudinal and seasonalvariations on retrieved TCWV is negligible (maximum value0.6 %), and that the systematic effect of CO2 and long-termCFCs variations over the missions are even smaller (0.07 %).

A5 Noise

Finally we analyse the impact of the measurement noise onthe retrieved TCWV (see last column of Table A1). The mea-surement noise was simulated by applying a random pertur-bation of ±0.037 K to the BTs of the two channels of ATSR-



2 and AATSR and a perturbation of ±0.1 K on the ATSR-1channels (Smith et al., 2012). For each instrument, we havegenerated 1000 values and we have evaluated the standarddeviation of the obtained TCWV that we report in absoluteand percentage values in Table A1. The standard deviation ismaximum at the poles, as expected, since there the TCWVand S /N ratio are lower. For AIRWAVEv2 we get 18 % forATSR-1 and 6 % for ATSR-2 and AATSR in the worst case.Note that AIRWAVEv2 approach has, for all scenarios, bet-ter performance with respect to AIRWAVEv1 (see Casadio etal., 2016).



TableA

1.Impactofdifferenterrorsources

onA

IRW

AVE

v2dataset.

Instrument

ScenarioTem

peratureH

2 Ow

indH

NO

3C

FC-11

CFC

-12C

O2

CFC

-11C

FC-12

CO

2N

oiseprofile

andSST

profile(25

ms−

1)trends

trendstrends

kgm−

2kg

m−

2kg

m−

2kg

m−

2kg

m−

2kg

m−

2kg

m−

2kg

m−

2kg

m−

2kg

m−

2kg

m−

2

AT

SR-1

Equatorial

3.20.5

−0.003

0.20.07

−0.07

0.003−

0.0060.014

0.014.8

(6.1%

)(1.0

%)

(−0.01

%)

(0.4%

)(0.14

%)

(−0.16

%)

(0.006%

)(−

0.01%

)(0.03

%)

(0.03%

)(9

%)

Midlatitude

1.20.1

0.2144.8

(4.2%

)(0.5

%)

(0.69%

)(16

%)

Polar0.3

0.10.448

3.0(1.6

%)

(0.5%

)(2.70

%)

(18%

)

AT

SR-2

Equatorial

3.50.5

−0.008

0.150.05

−0.05

−0.003

0.0150.009

−0.02

1.6(6.6

%)

(1.0%

)(−

0.01%

)(0.3

%)

(0.1%

)(−

0.1%

)(−

0.006%

)(0.03

%)

(0.02%

)(−

0.05%

)(3.1

%)

Midlatitude

1.40.2

0.2301.6

(5.1%

)(0.6

%)

(0.74%

)(5.2

%)

Polar0.4

0.10.467

1.0(2.0

%)

(0.5%

)(2.8

%)

(6.1%

)

AA

TSR

Equatorial

3.50.5

−0.012

0.280.08

−0.07

0.0060.01

0.01−

0.031.4

(6.7%

)(1.0

%)

(−0.02

%)

(0.56%

)(0.17

%)

(−0.15

%)

(0.01%

)(0.02

%)

(0.02%

)(−

0.07%

)(2.8

%)

Midlatitude

1.50.2

0.2041.5

(5.4%

)(0.6

%)

(0.66%

)(4.8

%)

Polar0.4

0.10.447

0.9(2.2

%)

(0.5%

)(2.7

%)

(5.6%

)



Author contributions. EC, EP, SC, BMD and BB contribute to thealgortithm development with ideas, comments and suggestions.ADR performed the tests used for the sistematic and random errorsquantifications. EP contribute to the validation of AIRWAVEv2.ECand EP wrote the manuscript. All authors contributed to interpreta-tions, read and commented on the manuscript.

Competing interests. The authors declare that they have no conflictof interest.

Acknowledgements. This work has been performed under the ESA-ESRIN contract no. 4000108531/13/I-NB. The authors gratefullyacknowledge ECMWF for SST data and the GlobVapour projectfor providing the combined MERIS+SSM/I water vapour product.SSM/I and SSM/IS data are produced by Remote Sensing Systems.Data are available at http://www.remss.com/missions/ssmi (lastaccess: 20 December 2018).

Edited by: Helen WordenReviewed by: two anonymous referees

References

Allan, R. P., Liu, C., Zahn, M., Lavers, D., Koukouva-gias, E., and Bodas-Salcedo, A.: Physically Consistent Re-sponses of the Global Atmospheric Hydrological Cycle inModels and Observations, Surv. Geophys., 35, 533–552,https://doi.org/10.1007/s10712-012-9213-z, 2014.

Aoki, S., Nakazawa, T., Machida, T., Sugawara, S., Morimoto, S.,Hashida, G., Yamanouchi, T., Kawamura, K., and Honda,H.:Carbon dioxide variations in the stratosphere over Japan, Scan-dinavia and Antarctica, Tellus B, 55, 178–186, 2003.

ARSA (Analyzed RadioSoundings Archive), available at: http://ara.abct.lmd.polytechnique.fr/index.php?page=arsa, last access: 17March 2018.

Berk, A., Acharya, P. K., Bernstein, L. S., Anderson, G. P., Lewis,P., Chetwynd, J. H., and Hoke, M. L.: Band model method formodeling atmospheric propagation at arbitrarily fine spectral res-olution, US Patent no. 7433806, 2008.

Casadio, S., Castelli, E., Papandrea, E., Dinelli, B. M., Pisacane,G., and Bojkov, B.: Total column water vapour from along trackscanning radiometer series using thermal infrared dual viewocean cloud free measurements: The Advanced Infra-Red WA-ter Vapour Estimator (AIRWAVE) algorithm, Remote Sens. En-viron., 172, 1–14, 2016.

Castelli, E., Papandrea, E., Valeri, M., Greco F. P., Ventrucci, M.,Casadio, S., and Dinelli, B. M.: ITCZ trend analysis via GeodesicP-spline smoothing of the AIRWAVE TCWV and cloud fre-quency datasets, Atmos. Res., 214, 228–238, 2018.

Delderfield, J., Llewellyn-Jones, D. T., Bernard, R., de Javel, Y.,Williamson, E. J., Mason, I., Pick, D. R., and Barton, I. J.: TheAlong Track Scanning Radiometer (ATSR) for ERS-1, Proc.SPIE, 589, 114–120, 1986.

Diedrich, H., Preusker, R., Lindstrot, R., and Fischer, J.: Retrievalof daytime total columnar water vapour from MODIS mea-

surements over land surfaces, Atmos. Meas. Tech., 8, 823–836,https://doi.org/10.5194/amt-8-823-2015, 2015.

Donlon, C., Berruti, B., Buongiorno, A., Ferreira, M.-H., Fémé-nias, P., Frerick, J., Goryl, P., Klein, U., Laur, H., Mavro-cordatos, C., Nieke, J., Rebhan, H., Seitz, B., Stroede, J., andSciarra, R.: The Global Monitoring for Environment and Secu-rity (GMES) Sentinel-3 mission, Remote Sens. Environ., 120,27–57, https://doi.org/10.1016/j.rse.2011.07.024, 2012.

Embury, O., Merchant, C., and Filipiak, M.: Refractive indices(500–3500 cm−1) and emissivity (600–3350 cm−1) of pure wa-ter and seawater, https://doi.org/10.7488/ds/162, 2008.

GEWEX G-VAP: GEWEX water vapour assessment, available at:http://gewex-vap.org/, last access: 17 March 2018.

Global Monitoring Division: Trends in atmospheric Carbon diox-ide, https://www.esrl.noaa.gov/gmd/ccgg/trends/data.html, lastaccess: 21 December 2018.

Li, Z.-L., Jia, L., Su, Z., Wan, Z., and Zhang, R.: A new approachfor retrieving precipitable water from ATSR2 split-window chan-nel data over land area, Int. J. Remote Sens., 24, 5095–5117,https://doi.org/10.1080/0143116031000096014, 2003.

Lindstrot, R., Stengel, M., Schröder, M., Schneider, N., Preusker,R., and Fischer, J.: Combined SSM/I and MERIS Water VapourProducts from the ESA GlobVapour project, in: AGU Fall Meet-ing Abstracts, Vol. 1, p. 0217, 2010.

Lindstrot, R., Preusker, R., Diedrich, H., Doppler, L., Bennartz, R.,and Fischer, J.: 1D-Var retrieval of daytime total columnar watervapour from MERIS measurements, Atmos. Meas. Tech., 5, 631–646, https://doi.org/10.5194/amt-5-631-2012, 2012.

Lindstrot, R., Stengel, M., Schröder, M., Fischer, J., Preusker,R., Schneider, N., Steenbergen, T., and Bojkov, B. R.: Aglobal climatology of total columnar water vapour fromSSM/I and MERIS, Earth Syst. Sci. Data, 6, 221–233,https://doi.org/10.5194/essd-6-221-2014, 2014.

Minschwaner, K., Hoffmann, L., Brown, A., Riese, M., Müller, R.,and Bernath, P. F.: Stratospheric loss and atmospheric lifetimesof CFC-11 and CFC-12 derived from satellite observations, At-mos. Chem. Phys., 13, 4253–4263, https://doi.org/10.5194/acp-13-4253-2013, 2013.

Papandrea, E., Casadio, S., De Grandis, E., Castelli, E., Dinelli, B.M., and Bojkov, B.: Validation of the Advanced Infra-Red WaterVapour Estimator (AIRWAVE) Total Column Water Vapour us-ing Satellite and Radiosonde products, Ann. Geophys., 61, 1–8,https://doi.org/10.4401/ag-7524, 2018.

Remedios, J. J., Leigh, R. J., Waterfall, A. M., Moore, D.P., Sembhi, H., Parkes, I., Greenhough, J., Chipperfield,M. P., and Hauglustaine, D.: MIPAS reference atmospheresand comparisons to V4.61/V4.62 MIPAS level 2 geophysi-cal data sets, Atmos. Chem. Phys. Discuss., 7, 9973–10017,https://doi.org/10.5194/acpd-7-9973-2007, 2007.

Rothman, L. S., Gordon, I. E., Barbe, A., Benner, D., Chris,Bernath, P. E., Birk, M., Boudon, V., Brown, L. R., Campar-gue, A., Champion, J.-P., Chance, K., Coudert, L. H., Dana,V., Devi, V. M., Fally, S., Flaud, J.-M., Gamache, R. R., Gold-man, A., Jacquemart, D., Kleiner, I., Lacome, N., Lafferty, W.J., Mandin, J.-Y., Massie, S. T., Mikhailenko, S. N., Miller,C. E., Moazzen-Ahmadi, N., Naumenko, O. V., Nikitin, A.V., Orphal, J., Perevalov, V. I., Perrin, A., Predoi-Cross, A.,Rinsland, C. P., Rotger, M., Simeckova, M., Smith, M. A. H.,Sung, K., Tashkun, S. A., Tennyson, J., Toth, R. A., Vandaele,


http://www.remss.com/missions/ssmi

https://doi.org/10.1007/s10712-012-9213-z

http://ara.abct.lmd.polytechnique.fr/index.php?page=arsa

http://ara.abct.lmd.polytechnique.fr/index.php?page=arsa

https://doi.org/10.5194/amt-8-823-2015

https://doi.org/10.1016/j.rse.2011.07.024

https://doi.org/10.7488/ds/162

http://gewex-vap.org/

https://www.esrl.noaa.gov/gmd/ccgg/trends/data.html

https://doi.org/10.1080/0143116031000096014

https://doi.org/10.5194/amt-5-631-2012

https://doi.org/10.5194/essd-6-221-2014

https://doi.org/10.5194/acp-13-4253-2013

https://doi.org/10.5194/acp-13-4253-2013

https://doi.org/10.4401/ag-7524

https://doi.org/10.5194/acpd-7-9973-2007


A. C., and Vander Auwera, J.: The HITRAN 2008 molecularspectroscopic database, J. Quant. Spectrosc. Ra., 110, 533–572,https://doi.org/10.1016/j.jqsrt.2009.02.013, 2009.

Schroder, M., Lockhoff, M., Fell, F., Forsythe, J., Trent, T., Ben-nartz, R., Borbas, E., Bosilovich, M. G., Castelli, E., Hers-bach, H., Kachi, M., Kobayashi, S., Kursinski, E. R., Loy-ola, D., Mears, C., Preusker, R., Rossow, W. B., and Saha, S.:The GEWEX Water Vapor Assessment archive of water vapourproducts from satellite observations and reanalyses, Earth Syst.Sci. Data, 10, 1093–1117, https://doi.org/10.5194/essd-10-1093-2018, 2018.

Seemann, S., Li, J., Menzel, W. P., and Gumley, L.: Operationalretrieval of atmospheric temperature, moisture, and ozone fromMODIS infrared radiances, J. Appl. Meteorol., 42, 1072–1091,2003.

Smith, D., Mutlow, C., Delderfield, J., Watkins, B., and Mason, G.:ATSR infrared radiometric calibration and in-orbit performance,Remote Sens. Environ., 116, 4–16, 2012.

Sobrino, J. A., Coll, C., and Caselles, V.: Atmospheric correctionsfor land surface temperature using AVHRR channel 4 and 5, Re-mote Sens. Environ., 38, 19–34, 1991.

Wentz, F. J.: A well-calibrated ocean algorithm for SSM/I, J. Geo-phys. Res., 102, 8703–8718, 1997.


https://doi.org/10.1016/j.jqsrt.2009.02.013



the advanced infra-red water vapour estimator (airwave...

Documents